Development of Integrating Data Mining System based on Information Base

多媒体数据挖掘是对多媒体对象进行数据挖掘，从多媒体对象或整体对象集合中发现模式。本研究致力于知识发现技术和应用的一些发展，特别是与大型图像和视频数据收集有关，如卫星图像和多维医学数据。主要论文“从时间序列图像中挖掘运动物体及其在卫星气象图像中的应用”通过具体的例子解决了挖掘视频序列或场景中运动物体的问题。我们考虑静止图像的集合，特别是卫星图像，但在这些图像中跟踪运动的物体，考虑图像序列作为帧。我们采用了一种两阶段自组织映射，将包含相似运动物体的图像聚类，从而识别出突出的物体，然后按时间顺序跟踪图像。将该框架应用于气象卫星图像，对季节变化趋势进行检测和分类。

Multimedia data mining is data mining from multimedia objects, the discovery of patterns from within multimedia objects or a collection of objects as a whole. This research was devoted to some developments in knowledge discovery techniques and application specifically pertaining to large image and video data collections such as satellite images and multi-dimensional medical data. The main paper "Mining of Moving Objects from Time Series Images and its Application to Satellite Weather Imagery'" addresses the issue of mining video sequences or moving objects in a scene with concrete examples. We consider collections of still images, in particular satellite images, but track moving objects within these images considering sequences of images as frames. We adapt a two-stage self-organizing map to cluster images in groups containing similar moving objects, thus identifying prominent objects which are then tracked in the sequence of chronologically ordered images. The framework was applied to weather satellite images to detect and classify seasonal variation tendencies.